The National Police of Colombia (NPC), in the development of educational programs that use canine teams (dog-handler), uses fundamental selection tests to determine which animals are fit or not to start training, continue it and finish their certification. Therefore, the objectives of this research are: 1) To propose a set of ethological indicators to evaluate the substance detector dog, based on an empirical field test based on instruments; and 2) Determine the qualification criteria for the substance detector dog during the association, empowerment and certification process for each of the evaluated tests. This allows the instruments to be validated to be reliable, and to predict the dogs that pass or fail the behavioral evaluations, contributing to the development of academic programs. To achieve the purpose of the study, a univariate analysis was carried out using 2x2 contingency tables, estimating the sensitivity and specificity for each of the tests performed on the canine detectors of narcotic and explosive substances (n = 549); determine the predictive values of the tests: Instrument No 1 (test-retest), Instrument No 2 (enhancement and association) and Instrument No 3 (final certification). In order to establish the level of agreement with the evaluators (Cohen's Kappa) by correlating the 17 individual and grouped behavioral variables to predict the "Fit" and "Not Fit" canines for police service. Among the main findings, high sensitivity and specificity is evidenced, with statistically significant results for most of the behavioral variables analyzed individually (P <0.05). Being "perseverance", the test that most influences the tests before training (Kappa of 1.0), in addition to an "almost perfect" level of agreement between the evaluators for themost tests, predicting Fit and Unfit canines to a high degree for the police service (Dogs not suitable in the final certification, n = 12; 3%). It is It is essential that academic programs that use binomials validate their tests, when establishing are the most representative variables for each evaluation in particular. Which improves and adjusts the instruments according to the needs of each institution or specialty of the canine service, reduces costs and improves quality standards, depending on the operational context to play for each binomial.